Pipeline

LLM enabled humanitarian data aggregation and visualisation pipeline.

Environment and requirements for main pipeline

• Create virtual environment for the porject: conda env create -f env/environment.yaml
• Activate virtual environment: conda activate llm-pipeline
• Import / update packages: pip install -r env/requirements.txt. NB: on Mac, you might need to run brew install gdal before using homebrew package manager, which will take quite some time so go grab a coffee. It is because of the geographical nature of the dashboard, which requires GDAL (a translator library for raster and vector geospatial data formats).

Environment variables

Environment variables are located in src/.env

• MISTRAL_API_KEY : your mistral API key to access the model
• NEW_API_KEY : please visit https://newsapi.org/, you can create an account and get an API key for free. Alternatively, if you don't want to do so, you could uncomment lines 176 to 179 in 'src/pipeline/get_api_data.py' and comment line 180.

Streamlit dashboard

• Make sure to be on the right conda virtual environment.
• Run streamlit run src/streamlit/dashboard.py.